Diagnostic accuracy of non-invasive methods detecting clinically significant portal hypertension in liver cirrhosis: a systematic review and meta-analysis.

Abstract

We attempted to investigate non-invasive techniques and their diagnostic performances for evaluating clinically significant portal hypertension. The systematic search was performed on PubMed, Embase, Scopus, and Web of Science TM core index databases before 13 December 2018 restricted to English language and human studies. Thirty-two studies were included, with total populations of 3,987. The overall pooled analysis was performed by bivariate random effect model, which revealed significantly higher sensitivity and specificity of 77.1% (95% confidence interval, 76.8-78.5%) and 80.1% (95% confidence interval, 78.2-81.9%), respectively; positive likelihood ratio (3.67), negative likelihood ratio (0.26); and diagnostic odd ratio (16.24). Additionally, the area under curve exhibited significant diagnostic accuracy of 0.871. However, notable heterogeneity existed in between studies (I2=87.1%), therefore, further subgroup analysis was performed. It demonstrated ultrasonography, elastography, biomarker, and computed tomography scan had a significant overall summary sensitivity (specificity) of 89.6% (78.9%), 81.7% (83.2%), 72.2% (76.8%), and 77.2% (81.2%), respectively. Moreover, the areas under curve values were significantly higher in elastography (0.906), followed by computed tomography scan (0.847), biomarker (0.825), and ultrasonography (0.803). In future, non-invasive techniques could be the future choice of investigations for screening and diagnosis of clinically significant portal hypertension in cirrhosis. However, standardization of diagnostic indices and their cut-off values in each non-invasive method needs to be addressed.